Classifying objects into categories is a critical cognitive process, for without it we would experience each object in the environment as unique and unconnected to our knowledge about any other object. "Prototype" theories of the categorization process assume that new objects are classified through retrieval of a summary or "prototype" representation of a category, while "exemplar" theories assume that classification involves retrieving stored representations of individual category members. Although the two views are very different in character, they have proven difficult to distinguish experimentally. The proposed research develops an on-line technique for assessing the type of information retrieved at the moment that a categorization decision is made. This technique avoids many of the problems inherent in the standard paradigms. Preliminary studies have supported the usefulness of the technique, and it appears to be applicable to investigating several central issues about categorization.